If you have a PDB file you can use COOT to mutate residues and real space refine. You can also download the electron density from the ESD through the COOT extensions. I suggest after mutating the residues to model the rotamer as closely as the experimentally determined structure (if there is an experimentally determined structure).

http://www.ccp4.ac.uk/index.php

for structure prediction: http://bioserv.rpbs.univ-paris-diderot.fr/services/PEP-FOLD/

for modifications:

1. Avogadro: http://avogadro.cc/wiki/Main_Page

2. VMD: http://www.ks.uiuc.edu/Training/Tutorials/vmd-molefacture/tutorial-Molefacture.pdf

I recommend chapter 15 (Stan Tsai : An Introduction to Computational Biochemistry) as a must read to learn about mutating, energy minimization and clash resolving.. etc.. 

as for structure prediction, I suggest SWISS-MODEL..

Most important is you have to have a hypothesis or a biological meaning of the modeling, otherwise you just end up with unpublishable data.. e.g. what this mutation will do to the protein and how it affects structure or function.. 

Thank You for such valuable help